On-farm broiler chicken welfare assessment using transect sampling reflects
environmental inputs and production outcomes
Neila BenSassi1☯, Judit Vas2☯, Guro VasdalID3, Xavier Avero´ s1, Inma Este´vez1,4, Ruth C. NewberryID2☯*
1 Department of Animal Production, Neiker-Tecnalia, Vitoria-Gasteiz, Spain, 2 Department of Animal and Aquacultural Sciences, Faculty of Biosciences, Norwegian University of Life Sciences,Ås, Norway, 3 Norwegian Meat and Poultry Research Centre, Oslo, Norway, 4 Ikerbasque, Basque Foundation for Science, Bilbao, Spain
☯These authors contributed equally to this work.
Abstract
To evaluate the utility of transect sampling for assessing animal welfare in large chicken flocks, we quantified relationships between environmental inputs, welfare problems
detected using transect sampling, and production outcomes. We hypothesised that environ- mental inputs including environmental complexity (i.e. number of environmental enrichment types provided), space allowance, underfloor heating (presence or absence), and photope- riod regimen (18 h continuous vs 16 h intermittent) would correspond to variations in welfare assessment findings, which would predict production outcomes. We conducted on-farm welfare assessment of Norwegian broiler flocks at approximately 28 days of age. We sam- pled four transects (rows between feeder and drinker lines) per flock to determine litter qual- ity and the proportions of chickens with compromised welfare as indicated by visual signs of walking difficulties, illness, skin wounds and small bird size. Production outcome measures included mortality, reasons for carcass rejection at slaughter, footpad dermatitis, growth rate, feed conversion and an integrated production index. Greater environmental complexity was associated with a reduction in skin wounds and total welfare problems on the farm, lower mortality, fewer rejections due to wounds and underweight birds, and fewer rejections overall. Higher space allowances within levels of environmental complexity were associated with fewer walking difficulties and welfare problems overall, a reduction in rejections due to wounds, and a higher growth rate and production index. Underfloor heating was associated with a reduction in rejections due to leg deformity, and intermittent light was associated with lower illness and skin wound rates on the farm, and lower mortality. Furthermore, fewer wel- fare problems and better litter quality on the farm were associated with fewer carcass rejec- tions at slaughter. Thus, data from transect sampling varied with environmental inputs and production outcomes, supporting the validity of transect sampling for practical, animal- based on-farm welfare assessment.
a1111111111 a1111111111 a1111111111 a1111111111 a1111111111
OPEN ACCESS
Citation: BenSassi N, Vas J, Vasdal G, Avero´s X, Este´vez I, Newberry RC (2019) On-farm broiler chicken welfare assessment using transect sampling reflects environmental inputs and production outcomes. PLoS ONE 14(4): e0214070.
https://doi.org/10.1371/journal.pone.0214070 Editor: Arda Yildirim, Tokat Gaziosmanpasa University, TURKEY
Received: July 20, 2018 Accepted: March 6, 2019 Published: April 17, 2019
Copyright:©2019 BenSassi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability Statement: The data are presented in the Supporting InformationS1 Appendix.
Funding: This research was financed by The Research Council of Norway (https://www.
forskningsradet.no/en/Home_page/
1177315753906), Grant No. 258881/E50 to RCN, and Spanish Ministry of Economy and
Competitiveness (http://www.mineco.gob.es/), Grant no. ERA-NET 291815 (ANIHWA) to IE, in relation to the European Union Animal Health and
Introduction
In modern animal agriculture, animals are kept in large groups, with flocks, schools, and herds numbering in the thousands. Keeping track of the welfare of individual animals in such large groups presents challenges. It is, thus, common to base animal welfare assessment on adher- ence to engineering standards (i.e.a prioriresource-based rules such as the provision of a cer- tain amount of space/animal). However, because animal welfare is about quality of life, animal welfare assessment has greater face validity when based on observation of animals than infer- ence based on resource provision (e.g. [1,2]). Assessment of living animals also has greater face validity than welfare assessment based ona posteriorireview of animal-based production records and slaughter plant health inspection outcomes. Yet, existing on-farm welfare assess- ment protocols (e.g. [3]) tend to be time-consuming [2,4,5], making them more suited for detailed research investigations than as practical industry management tools. In addition, ani- mals can be stressed by protocols that require them to be handled for close examination of their physical condition, which may affect results [4]. There is a need, therefore, for simple and efficient, low-stress tools for welfare assessment on the farm.
Practical welfare surveillance methods for use in large commercial poultry houses are par- ticularly needed. Poultry kept for meat production can be afflicted by a variety of welfare chal- lenges including wounds, infections, and cardiovascular and skeletal metabolic disorders that compromise normal development [6,7,8,9]. To assess the prevalence of such welfare prob- lems in a practical, low-stress but systematic manner, a transect sampling procedure has been developed [4] and validated against the “gold standard” of evaluating every bird in the flock [10]. According to the transect sampling method, an observer walks slowly along successive transects in the house, defined as rows between lines of feeders and drinkers. In each transect, birds located in the space immediately ahead of the observer are scanned as they move out of the path of the approaching observer. The observer records all birds within the transect that are observed to be clearly afflicted by specific welfare-relevant conditions as identified by the European Food Safety Authority Panel on Animal Health and Welfare [11]. For broiler chick- ens, these welfare “red flag” indicators include: lame, immobile, head, back and tail wounds, small, dirty, featherless, sick, terminally ill, and dead birds. To optimise surveillance time and minimise the risk of missing or recounting birds, only clearly evident “iceberg” cases that can be rapidly categorised are recorded, rather than attempting to score every bird on a graded scale of severity. Litter quality is also rated at three points along each transect. A free android smartphone application (i-WatchBroiler) has been designed for use in entering the data col- lected during transect walks in broiler houses [12].
Transect sampling is intended for implementation by stakeholders including farmers, veter- inarians, animal welfare auditors and advisors to obtain a quick but quantitative snapshot of the current welfare status of the flock. However, previous research indicates the potential for some variation in results from different observers and sampled transects [4,10,13], which could mask differences in welfare between flocks. It has not yet been established whether tran- sect sampling can reveal differences in flock welfare according to environmental conditions (e.g. environmental enrichment, space allowance, heating systems, lighting programmes), or offer a useful forecast of production outcomes.
Environmental enrichment refers to additions and modifications to the housing environ- ment that increase environmental complexity, stimulate species-specific behaviour and facili- tate biological functioning, thereby improving animal welfare [14,15]. Elevated resting surfaces such as perches, platforms and boxes can serve as environmental enrichment for broilers [16]. Use of such structures may strengthen muscles and joints and enhance the blood supply to the legs, wings and heart [17]. Their provision has been associated with greater tibial
Welfare Era-Net (ANIHWA) project “Integrated mobile broiler data—optimising broiler chicken management, health and welfare through use of integrated data”. NBS was supported by a Ph.D.
scholarship (BES-2014-070242) and travel grants (EEBB-I-17-12574 and EEBB-I-18-12937) from the Spanish Ministry of Economy and Competitiveness (http://www.mineco.gob.es/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
diaphysis diameter [18], a reduction in tibial dyschondroplasia [19], reduced severity of foot- pad dermatitis [20], lower mortality due to heat stress [21], and a lower heterophil to lympho- cyte ratio [22,23]. Peat moss is an environmental enrichment material that stimulates ground scratching and dustbathing behaviour [24,25,26]. A negative association between lameness score and dustbathing frequency has been reported [27], suggesting that peat could improve leg health. Moreover, peat can have beneficial effects on the digestive tract mucosa [28], and its use as a feed supplement has been associated with increased weight gain, and feed efficiency [29,30,31]. Bales of litter material provide enrichment by stimulating exploratory pecking and foraging behaviour as well as acting as platforms [32,33]. Collectively, increased environmen- tal complexity resulting from provision of multiple types of enrichment simultaneously may have additive welfare benefits.
Space allowance is also of relevance to broiler welfare and production outcomes. In differ- ent studies, increasing the space allowance from 0.044–0.083 m2/bird [34], 0.046–0.074 m2/ bird [35], or 0.066–0.500 m2/bird [36], reduced the prevalence or severity of footpad dermati- tis. An increase in space allowance from 0.044–0.083 m2/bird [34], or reduction in the stocking density from 44.8–15.9 kg/m2[37], also improved walking ability. In addition, more space/
bird has been associated with lower mortality, contact dermatitis and carcass bruising [38], and increased growth and feed efficiency (e.g. when comparing 30–35 kg/m2with 40 kg/m2 [35,39,40,41] and 50 kg/m2[42]).
Underfloor heating is reported to improve temperature uniformity and efficiency of tem- perature distribution in comparison to hood heating [43]. It may also reduce litter moisture and ammonia, resulting in reduced cardiovascular disease and ascites [44], leg problems [37], footpad dermatitis and feather dirtiness [45]. Underfloor heating has been associated with a reduction in footpad dermatitis in turkeys [46,47], and with lower mortality, greater weight gain and lower feed consumption in broilers [48].
The photoperiod regimen can affect body temperature and the immune system [49], and regulate feed intake [50] and physical activity [51]. Long photoperiods (20–24 h) have been associated with susceptibility to leg problems in broilers [37,52] and increased mortality and leg problems in Japanese quail [53]. An intermittent photoperiod regimen alternating between 2 h light and 2 h dark has been associated with less footpad dermatitis and higher body weights in broilers when compared to a short (8 h) continuous photoperiod [54]. Provision of at least 6 h of darkness/day is now required in the European Union, with at least one uninterrupted dark period of at least 4 h [55]. In Norway, the latter provision is limited to two uninterrupted dark periods of at least 4 h [56]. These requirements have resulted in two typical forms of pho- toperiod regimen in Norway, one comprising 18 h of continuous light and the other having two 4-h dark periods/day (i.e. 16 h intermittent light).
In the current study, we used the transect sampling method to collect data from Norwegian commercial broiler flocks. Our aim was to investigate relationships between environmental factors, transect data and production data (including health inspection findings from the slaughter line where every bird in the flock is evaluated). Based on previous reports, we hypothesised that environmental complexity (defined as the number of environmental enrich- ment types provided), space allowance, underfloor heating, and photoperiod regimen would be correlated with both transect and production findings. Specifically, we predicted that greater environmental complexity, greater space allowance, presence (vs absence) of under- floor heating, and 16 h intermittent (vs 18 h continuous) lighting would be associated with indications of improved welfare both on the farm and at slaughter. Further, we expected to find positive associations between on-farm welfare problems and causes of rejection at slaugh- ter, and a negative association between litter quality and footpad dermatitis. Additionally, we
examined the consistency of transect data between observers, and between transect locations within the house (left vs right side, beside wall vs more central).
Materials and methods Ethics statement
The study was conducted between February and May, 2017, on 15 farms located in southeast Norway. All farms belonged to the same cooperative, and functioned in accordance with Nor- wegian animal welfare legislation governing poultry production [56]. Farm owners gave their consent to participate in the research, participation was voluntary, and no personal details were collected. No biological samples were collected for research purposes. Because no experi- mental manipulations were made and observations were non-invasive, the study did not require approval of animal use by the Norwegian Food Safety Authority [57].
Environmental inputs
At each farm, we evaluated two consecutive Ross 308 mixed sex flocks kept in the same house (Table 1). All houses were well-insulated, with concrete floors and automatic drinkers, feeders and ventilation systems. Ten houses had underfloor heating. Houses were thoroughly cleaned and supplied with a thin layer of fresh litter material (generally softwood shavings) before placement of each flock. Initial flock size ranged from 9,600–34,050 broilers (mean±SE, 19,480±809) and initial space allowance ranged from 0.056–0.073 m2/bird. Artificial lighting was provided by LED lights, with either a single 6-h dark period or two 4-h dark periods daily from 7 days of age until 3 days before slaughter. Farmers checked their flocks at least twice daily. They removed any birds found dead, humanely culled any moribund or severely dis- abled birds, and kept records on these numbers. On the day of slaughter (between 32 and 35 days of age), the mean stocking density (±SE) was 32.08±0.63 kg/m2.
The cooperative had a recommended environmental enrichment programme, which involved providing boxes for perching by 7 days of age (1 box/50 m2), and peat (10 l/50 m2) and wood shavings bales (1 half-bale/100 m2) at 7, 14, 21 and 28 days of age. Farmers supplied these enrichments to varying degrees in the different flocks (Table 1). The boxes were either cardboard or plastic, ranging from about 0.2–0.25 m high and with an upper surface area of about 0.2–0.3 m2. Some boxes had openings allowing birds to go under them when young, and some were stacked in pyramids. The boxes remained in the house throughout rearing. Peat was provided over the whole floor (as litter), loose in piles, contained in low surrounds, or as bales (200 l bales, presented whole or cut in half). Wood shavings bales (25 kg) were cut in half and presented with or without removal of their plastic wrapping. Because the added peat and wood shavings bales became integrated into the litter, they were renewed weekly.
On-farm welfare assessment
Two trained observers visited each flock once at 26–30 days of age, shortly before slaughter when welfare problems were most evident. Before starting the data collection, we determined the dimensions of the house (mean±SE, 1284±53 m2), and width of each transect
(mean±SE, 2.06±0.04 m). Transects were defined as wall or central transects. Wall transects comprised the area demarcated by a side wall and the nearest feeder or drinker line (whichever was closest to the wall, typically a drinker line), extending the length of the house from one end wall to the other end wall. Central transects comprised the area bounded by adjacent feeder and/or drinker lines (typically one of each), extending the length of the house. Transects
were numbered consecutively starting with the wall transect on the left side of the house, as viewed when standing at the end of the house closest to the entrance door.
On farm visits, each observer assessed the prevalence of the welfare indicators (Table 2) within one central transect and one wall transect, for a total of four transects/flock. One observer sampled two transects on the left side of the house, walking up one transect and returning down the other, and the second observer sampled two transects on the right side of the house. While one observer sampled a wall transect, the other simultaneously sampled a central transect, and vice versa. We randomised the side of the house evaluated by each observer, and alternated the order of observing wall and central transects. Both wall transects were included in the flock assessment because observations of Spanish broiler flocks indicated that immobile, small, sick, dirty and/or dead birds were more likely to be found in wall than central transects [4,13]. We selected the two observed central transects pseudo-randomly,
Table 1. Environmental provisions to each flock.
Farm Under-floor heating
Lighting regimen (18 h continuous vs 16 h intermittent)
Flock Space allowance (m2/chick started)
Environmental enrichment type (X indicates provision) Boxes Peat Wood shavings
bales
1 Yes 18 h 1 0.070 X X X
2 0.061 X
2 Yes 18 h 1 0.066 X X X
2 0.073 X X X
3 No 16 h 1 0.057
2 0.073 X X X
4 Yes 18 h 1 0.056
2 0.072 X X X
5 Yes 16 h 1 0.057
2 0.072 X X X
6 No 18 h 1 0.057 X X X
2 0.072 X X X
7 Yes 16 h 1 0.068 X X X
2 0.058
8 Yes 18 h 1 0.061 X X X
2 0.074 X X X
9 Yes 18 h 1 0.063 X
2 0.071 X X
10 No 16 h 1 0.072 X X X
2 0.060 X X
11 No 16 h 1 0.062
2 0.071 X X X
12 No 18 h 1 0.069 X X X
2 0.063 X
13 Yes 16 h 11 0.058 X X
2 0.060 X
14 Yes 18 h 1 0.060 X X
2 0.066 X
15 Yes 18 h 1 0.061 X X
2 0.063
1On-farm welfare assessment data missing from this flock.
https://doi.org/10.1371/journal.pone.0214070.t001
avoiding contiguous transects to minimise double counting of the same birds, and any atypi- cally wide (>3 m) or narrow (<1 m) transects. Evaluation of two transects took an average of 30–35 min depending on house length. Birds were assigned to the welfare indicator best describing their condition based on rapid visual assessment.
We observed no dirty or featherless birds in the flocks visited. Due to low numbers, we assigned the transect counts for the remaining indicators to four broader categories: walking difficulties, illness, skin wounds, or small (Table 2). The counts in each welfare category were summed across the four assessed transects and expressed as a proportion of the total number of birds estimated to be present in those four transects. This denominator was calculated based on the total number of birds present in the house on the assessment day and the dimensions of the transects, assuming a uniform distribution of birds across the house. We also calculated an overall welfare problems index (i.e. summed counts across all categories as a proportion of the estimated number of birds in the observed transects). We evaluated litter quality at the begin- ning, middle, and end of each walked transect on a 5-point scale, from 0 (dry and loose litter) to 4 (caked litter) based on the Welfare Quality [3] protocol for poultry, and calculated the average litter score/flock.
Production outcomes
Flocks were slaughtered at a mean age of 33.6 days, all at the same slaughter plant following 2-phase CO2gas stunning. Production data on each flock were provided by the farmers and the slaughter plant. We calculated total mortality on the farm up to the day of slaughter as [(found dead + culled)/number of chicks started]. Reasons for carcass rejection were routinely recorded by health inspection personnel stationed along the slaughter line. They recorded the
Table 2. Ethogram of broiler welfare problems recorded during transect sampling, and subsequently pooled categories.
Indicator Description Category
Lame Walks with obviously uneven strides or unsteady steps. May exhibit outward or inward twisting of one or both legs leading to severe limping. Lameness is clearly advanced rather than in early stages.
Walking difficulties Immobile Does not move away when approached or moves by propping on wings or
crawling. If gently nudged, moves with difficulty, no more than three steps before sitting down again.
Sick Signs of impaired health, including small and/or pale comb, red, watery or closed eyes, retracted neck and disarranged/raised feathers. Usually found in a resting position. Includes wry neck.
Illness
Terminally ill
Lying with head resting on ground or lying on back, with signs of being close to death (e.g. laboured breathing, half-closed eyes). Excludes panting related to heat stress.
Dead No signs of life.
Head wounds
Skin scratches on head or neck indicated by the presence of fresh or dried blood/
scabs visible from 1–2 m away.
Skin wounds Back
wounds
Skin scratches on back (between neck and tail) and/or wings indicated by the presence of fresh or dried blood/scabs visible from 1–2 m away.
Tail wounds Skin scratches around tail and hips indicated by the presence of fresh or dried blood/scabs visible from 1–2 m away.
Small Stunted growth. Approximately half average size of flock mates. May have yellow downy feathers, especially on head.
Small Dirty Extensive dark staining of body sides, wings, chest, back, and/or tail feathers due
to prolonged contact with wet litter. Excludes light soiling or discolouration of feathers caused by dust, peat or excrement.
Not observed
Featherless Lacking feathers on majority of back and wings. Excludes moulting. Not observed https://doi.org/10.1371/journal.pone.0214070.t002
primary reason for rejection of each bird though multiple reasons could exist. We categorised these reasons as: perosis (any pronounced leg deformities), illness (sum of liver disease, heart disease, ascites, persistent egg yolk, and discolouration/suspicious smell), wounds (scratches, bruises, hematomas, fractures and dislocations), and underweight (below marketable weight).
We expressed the numbers rejected in each category, and total number rejected, as a propor- tion of the total number of birds slaughtered. Footpad dermatitis was evaluated by slaughter- house personnel according to standard procedure for Norwegian flocks, whereby 100 feet/
flock were assessed on a 3-point scale (0 = no lesions, 1 = mild lesions, 2 = severe lesions), and points were summed to give a flock score ranging from 0–200. Further flock data included growth rate [mean g eviscerated carcass weight/days of age at slaughter], and the feed conver- sion ratio [total kg feed provided/((number slaughtered−number rejected)�mean kg eviscer- ated carcass weight)]. We also calculated an integrated production index value for each flock [mean g eviscerated carcass weight�(number slaughtered−number rejected)/(days of age at slaughter�number of chicks started)].
Statistical analysis
We analysed all data using SAS 9.4 (SAS Institute, NC, USA). Associations of the four environ- mental inputs (environmental complexity, space allowance, underfloor heating, and photope- riod regimen) with each on-farm welfare assessment variable and each production outcome variable were investigated using generalised linear mixed models (GLIMMIX procedure).
Environmental complexity (i.e. number of environmental enrichment types provided) was treated as a continuous variable ranging from 0–3 to explore linear trends irrespective of the specific combinations of enrichment materials used. This approach recognised the underlying continuity of complexity despite imprecise quantification. Due to collinearity between space allowance and environmental complexity, the residuals of space allowance regressed on envi- ronmental complexity were included in the model as a continuous variable describing the vari- ation in space allowance around the regression line at each level of environmental complexity.
Underfloor heating (absence vs presence) and photoperiod regimen (18 h continuous vs 16 h intermittent) were categorical factors. We used additional GLIMMIX models to estimate associations of the on-farm welfare assessment variables (1) walking difficulties, illness, skin wounds, and small birds, (2) the welfare problems index, and (3) litter score, with the produc- tion outcomes (total mortality, reasons for rejection at slaughter, total rejections, footpad der- matitis score, growth rate, feed conversion ratio, production index). Farm was included as a random effect in all models (seeS1 Appendixfor model specification details and covariance estimates for farm).
Response variables comprising counts expressed as proportions were analysed according to the binomial distribution with logit link, maximum likelihood estimation and Laplace likeli- hood approximation. Because flock footpad dermatitis scores were heavily right-skewed, with a majority of flocks receiving a score of 1, we compared flocks receiving scores of 1 vs>1 based on the binary distribution with logit link. We analysed the remaining response variables (mean litter score, growth rate, feed conversion, production index) according to the gamma distribution with log link and residual pseudo-likelihood estimation (seeS1 Appendixfor details). We applied the inverse link to back-transform estimated values (continuous factors) and least squares means (categorical factors) to their original scale for graphical presentation.
The absence of underfloor heating, and a continuous photoperiod of 18 h, served as the refer- ence levels for least squares means estimation.
We also evaluated the degree of agreement in findings on the proportion of birds with walk- ing difficulties, illness, skin wounds, and small birds between pairs of transects within flocks.
We compared the differences between the pairs of transects that were (1) assessed by different observers, (2) located on the left vs right side of the house and (3) located in wall vs central transects, using the Wilcoxon signed-ranks test in the UNIVARIATE procedure.
Results
Descriptive data
The welfare problem indicators occurred at low levels (Table 3;S1 Appendix), with lameness contributing most to the welfare problems index. Skin wounds were most common in the tail region. The litter scores at sampled locations varied from 0–2, with no scores of 3 or 4 being recorded. Culling by the farmer accounted for about 30% of the mortality. Ascites and liver disease were the most common reasons for rejection at slaughter, resulting in the pooled illness category accounting for the majority of rejections. Footpad dermatitis occurred at low levels, with flock scores ranging from 1–13 of a possible 200.
Table 3. Prevalence of welfare problems detected by transect sampling during on-farm welfare assessment, and production outcomes.
Welfare indicator1 Mean SE Production outcome2 Mean SE
Lame (%) 0.22 0.02 Mortality (%)3 3.59 0.35
Immobile (%) 0.07 0.01 Culled (% of mortality)4 30.31 3.91
Walking difficulties (%)5 0.29 0.03 Rejection due to perosis (%) 0.01 <0.01
Sick (%) 0.03 <0.01 Rejection due to liver disease (%) 0.35 <0.01
Terminally ill (%) <0.01 <0.01 Rejection due to heart disease (%) 0.07 0.01
Dead (%) 0.04 0.01 Rejection due to ascites (%) 0.71 0.06
Illness (%)6 0.07 0.01 Rejection due to persistent egg yolk (%) 0.01 <0.01
Head wounds (%) 0.01 <0.01 Rejection due to discolouration/smell (%) 0.05 0.01
Back wounds (%) <0.01 <0.01 Rejection due to illness (%)7 1.19 0.11
Tail wounds (%) 0.08 0.01 Rejection due to wounds (%) 0.14 0.04
Skin wounds (%)8 0.08 0.01 Rejection because underweight (%) 0.20 0.08
Small (%) 0.12 0.04 Total rejections (%)9 1.66 0.15
Dirty (%) 0 0 Footpad score10 2.37 0.46
Featherless (%) 0 0 Growth rate (g/day)11 43.20 0.42
Welfare problems index (%)12 0.57 0.06 Feed conversion (ratio)13 2.22 0.02
Litter score14 1.17 0.06 Production index (g/day)15 40.86 0.42
1On-farm welfare indicator data from transect sampling on 15 farms (2 flocks/farm; n = 29 flocks). Counts expressed as % of estimated number of birds in four walked transects per flock.
2Flock production data (n = 30 flocks). Reasons for rejection counts expressed as % of total number of birds slaughtered.
3[(Found dead + culled)/number of chicks started] up to day of slaughter, expressed as %.
4Number culled, as a % of mortality up to day of visit (n = 26 flocks due to missing data).
5Includes lame and immobile.
6Includes sick, terminally ill and dead.
7Includes liver disease, heart disease, ascites, persistent egg yolk and discolouration/suspicious smell.
8Includes head, back and tail wounds.
9Includes perosis/leg deformity, rejection due to illness, rejection due to wounds and rejection because underweight.
10100 feet/flock scored on 3-point scale (0 = no lesions, 1 = mild lesions, 2 = severe lesions), giving a maximum possible flock score of 200.
11[Mean g eviscerated carcass weight/days of age at slaughter].
12Includes lame, immobile, sick, terminally ill, dead, head, back, and tail wounds, and small.
13[Total kg feed provided to flock/((number slaughtered−number rejected)�mean kg eviscerated carcass weight)].
14Scored from 0 (dry, loose litter) to 4 (caked litter) in three locations / transect.
15[Mean g eviscerated carcass weight�(number slaughtered−number rejected)/(days of age at slaughter�number of chicks started)].
https://doi.org/10.1371/journal.pone.0214070.t003
Associations of environmental inputs with welfare assessment findings and production outcomes
With increasing environmental complexity (Table 4), we detected fewer birds with skin wounds during the transect walks (P = 0.004;Fig 1A), and the overall welfare problems index was lower (P = 0.002;Fig 1B). Increasing environmental complexity was associated with lower mortality (P<0.001;Fig 1C), a lower proportion of rejections due to wounds (P<0.001;Fig 2A) and underweight birds (P = 0.002;Fig 2B), and a lower overall rejection rate (P<0.001;
Fig 2C).
As space allowance residuals increased (Table 4), fewer birds with walking difficulties were detected (P<0.001;Fig 3A) and the welfare problems index was lower (P = 0.006;Fig 3B).
Higher space allowance residuals were also associated with a lower proportion of slaughter rejections due to wounds (P<0.001;Fig 4A), a higher growth rate (P = 0.006;Fig 4B), and a higher production index overall (P = 0.025;Fig 4C).
The significant associations of environmental complexity and space allowance with welfare assessment findings and production outcomes had low to modest r2values ranging from 0.007 to 0.480 (Figs1–4).
The presence of underfloor heating (Table 5) was associated with fewer rejections due to perosis (P = 0.037;Fig 5A). The 16 h intermittent photoperiod regimen (Table 5) was associ- ated with lower rates of illness (P = 0.015;Fig 5B) and skin wounds (P = 0.026;Fig 5C) on the farm than the 18 h continuous photoperiod regimen, as well as lower mortality (P = 0.022;
Fig 5D).
Table 4. Regression coefficient estimates for associations of environmental complexity and space allowance with welfare problems detected by transect sampling, and production outcomes.
Variable Environmental complexity1 Residuals of space allowance2
Mean SE Lower 95% CL Upper 95% CL F3 P Mean SE Lower 95% CL Upper 95% CL F3 P
On-farm welfare assessment indicators
Walking difficulties -0.09 0.04 -0.18 -0.01 4.50 0.055 -57.95 11.09 -82.11 -33.79 27.31 <0.001
Illness -0.04 0.08 -0.20 0.13 0.24 0.632 5.77 19.48 -36.66 48.21 0.09 0.772
Skin wounds -0.26 0.08 -0.43 -0.10 12.46 0.004 -31.27 20.41 -75.73 13.19 2.35 0.151
Small -0.09 0.07 -0.23 0.05 1.94 0.189 34.11 17.25 -3.49 71.70 3.91 0.072
Welfare problems index -0.12 0.03 -0.19 -0.06 16.29 0.002 -26.79 7.93 -44.07 -9.51 11.41 0.006
Litter score -0.09 0.05 -0.20 0.03 2.62 0.131 3.16 14.32 -28.03 34.35 0.05 0.829
Production outcomes
Mortality -0.05 0.01 -0.06 -0.03 35.81 <0.001 1.09 1.91 -3.03 5.21 0.33 0.577
Rejection due to perosis -0.19 0.15 -0.52 0.14 1.48 0.245 -56.67 38.87 -140.63 27.30 2.13 0.169
Rejection due to illness 0.03 0.01 <0.01 0.06 4.45 0.055 -2.92 3.50 -10.47 4.64 0.70 0.419
Rejection due to wounds -0.35 0.04 -0.44 -0.27 88.99 <0.001 -78.88 13.51 -108.05 -49.70 34.11 <0.001
Rejection because underweight -0.16 0.04 -0.25 -0.07 15.08 0.002 17.49 9.83 -3.74 38.72 3.17 0.099
Total rejections -0.06 0.01 -0.08 -0.04 29.35 <0.001 -1.28 2.96 -7.66 5.11 0.19 0.673
Footpad score (binary) 0.10 0.34 -0.64 0.84 0.09 0.774 121.44 92.39 -78.15 321.03 1.73 0.211
Growth rate -0.01 0.01 -0.02 0.01 2.17 0.164 4.89 1.50 1.65 8.12 10.63 0.006
Feed conversion <0.01 0.01 -0.01 0.01 0.01 0.927 -1.64 1.45 -4.78 1.50 1.27 0.280
Production index -0.01 0.01 -0.02 0.01 0.46 0.511 4.68 1.85 0.68 8.67 6.40 0.025
1Number of environmental enrichment types (boxes, peat, wood shavings bales) provided (0–3).
2Residuals of space allowance (m2/bird) regressed on environmental complexity.
3F1, 12for welfare assessment variables; F1, 13for production variables. SeeMethodsandTable 3footnotes for explanation of variables.
https://doi.org/10.1371/journal.pone.0214070.t004
Fig 1. Associations of environmental complexity with (A) skin wounds (r2= 0.183), (B) welfare problems index (r2= 0.164), and (C) mortality (r2= 0.074). Environmental complexity is based on the number of environmental enrichment types (boxes, peat, wood shavings bales) provided (from 0–3). Data points are back-transformed estimates. (A, B) Birds detected with skin wounds, and sum of birds detected with welfare problems (walking difficulties, illness, skin wounds, small size), as a % of the estimated number of birds in 4 assessed transects. (C) Number found dead and culled on the farm up to the day of slaughter as a % of number of chicks started.
https://doi.org/10.1371/journal.pone.0214070.g001
Fig 2. Associations of environmental complexity with rejections due to (A) wounds (r2= 0.206), and (B) underweight birds (r2= 0.053), and (C) total rejections (r2= 0.076). Environmental complexity is based on the number of environmental enrichment types (boxes, peat, wood shavings bales) provided (from 0–3). Data points are back-transformed estimates. (A, B, C) Carcasses rejected as a % of total number of slaughtered birds.
https://doi.org/10.1371/journal.pone.0214070.g002
Fig 3. Associations of space allowance (m2/bird) with (A) walking difficulties (r2= 0.480) and (B) welfare problems index (r2= 0.233). Data points are back-transformed estimates from analysis of residuals of space allowance regressed on environmental complexity. (A) Birds detected with walking difficulties as a % of estimated number of birds in 4 assessed transects. (B) Sum of birds detected with welfare problems (walking difficulties, illness, skin wounds, small size), as a % of the estimated number of birds in 4 assessed transects.
https://doi.org/10.1371/journal.pone.0214070.g003
Fig 4. Associations of space allowance (m2/bird) with (A) rejections due to wounds (r2= 0.151), (B) growth rate (r2= 0.007), and (C) production index (r2= 0.048). Data points are back-transformed estimates from analysis of residuals of space allowance regressed on environmental complexity. (A) Carcasses rejected due to wounds as a % of total slaughtered birds. (B) [Mean g eviscerated carcass weight/days of age at slaughter]. (C) [Mean g eviscerated carcass weight�(number slaughtered−number rejected)/(days of age at slaughter�number of chicks started)].
https://doi.org/10.1371/journal.pone.0214070.g004
Table5.Back-transformedleastsquaresmeansforassociationsofunderfloorheatingandphotoperiodregimenwithwelfareproblemsdetectedbytransectsamplingandproduction outcomes. VariableUnderfloorheatingPhotoperiodregimen WithoutWithF1P18hcontinuous16hintermittentF1P MeanSELower 95%CLUpper 95%CLMeanSELower 95%CLUpper 95%CLMeanSELower 95%CLUpper 95%CLMeanSELower 95%CLUpper 95%CL On-farmwelfareassessmentindicators Walking difficulties(%)0.230.030.170.310.290.030.230.371.660.2210.270.030.210.350.250.040.180.330.270.616 Illness(%)0.080.010.050.120.050.010.040.073.620.0810.090.010.070.120.050.010.030.078.110.015 Skinwounds(%)0.070.020.040.120.060.010.040.090.140.7170.100.020.070.140.040.010.030.086.470.026 Small(%)0.050.020.020.100.100.030.050.172.220.1620.080.020.040.140.060.020.030.130.180.678 Welfare problemsindex (%)
0.440.070.320.620.510.060.400.700.550.4720.570.070.440.740.400.060.290.563.100.104 Litterscore1.340.161.021.751.080.110.871.341.720.2141.130.110.911.411.270.160.971.660.440.518 Productionoutcomes Mortality(%)3.220.512.294.532.860.372.173.770.330.5783.980.503.035.232.310.371.633.266.710.022 Rejectiondueto perosis(%)0.01<0.010.010.02<0.01<0.01<0.010.015.410.0370.01<0.01<0.010.010.01<0.01<0.010.010.150.709 Rejectiondueto illness(%)1.340.210.961.870.980.120.751.282.330.1511.250.160.961.641.050.160.751.460.770.397 Rejectiondueto wounds(%)0.060.030.020.150.070.030.030.150.120.7390.090.030.040.180.050.020.020.130.800.387 Rejection because underweight(%)
0.080.030.030.190.090.030.040.180.050.8230.110.040.050.240.060.030.020.151.370.263 Totalrejections (%)1.650.261.172.331.350.171.021.780.920.3551.790.231.362.351.250.200.881.772.860.115 Footpadscore (binary)0.390.170.130.750.380.130.160.68<0.000.9620.510.140.240.770.280.150.080.651.020.331 Growthrate(g/ day)42.180.8240.4444.0043.680.6942.2245.191.800.20342.860.6841.4344.3542.990.8441.2144.840.010.914 Feedconversion (ratio)2.270.042.192.352.210.032.142.271.710.2142.230.032.162.292.250.042.172.330.180.676 Production index(g/day)39.930.8138.2241.7241.540.6840.0943.042.130.16840.190.6638.7941.6441.270.8439.5043.130.950.346 1F1,12forwelfareassessmentvariables;F1,13forproductionvariables.SeeMethodsandTable3footnotesforexplanationofvariables. https://doi.org/10.1371/journal.pone.0214070.t005
Relationships between welfare assessment findings and production outcomes
An increased prevalence of walking difficulties, illness and small size on the farm was associ- ated with increased mortality, and increased rejections due to illness and underweight birds, and increased total rejections at slaughter (P<0.05;Table 6). A higher prevalence of walking difficulties was also associated with increased rejections due to wounds (P<0.001). A higher welfare problems index on the farm was associated with higher mortality, rejections in the ill- ness, wounds, and underweight categories, and total rejections (P<0.001). Higher litter scores were associated with lower mortality, but higher rejections due to illness, wounds and under- weight birds, as well as total rejections (P<0.01;Table 6). Litter scores were not associated with footpad dermatitis scores in this study.
Fig 5. Associations of underfloor heating with (A) rejection due to perosis, and of photoperiod regimen on on-farm (B) illness, (C) skin wounds, and (D) mortality. Underfloor heating (UFH, without vs with). Photoperiod regimen (18 h continuous vs 16 h intermittent). Bars show back-transformed least squares means±SE (differences, P<0.05). (A) Carcasses rejected due to perosis (leg deformity) as a % of total number of slaughtered birds. (B, C) Birds detected with signs of illness, and skin wounds, as a % of the estimated number of birds in 4 assessed transects. (D) Number found dead and culled on the farm up to the day of slaughter as a % of number of chicks started.
https://doi.org/10.1371/journal.pone.0214070.g005
Table 6. Regression coefficient estimates for relationships between welfare problems detected by transect sampling, and production outcomes.
Production outcomes1 On-farm welfare assessment
indicators2
Mortality Reasons for rejection Total
rejected
Footpad score Growth rate Feed conversion Production index Perosis Illness Wounds Under-
weight
Walking difficulties Mean 162.53 254.75 38.82 457.45 146.61 78.23 -121.23 0.30 -4.44 -2.74
SE 10.01 183.17 16.51 53.87 49.68 14.17 306.29 6.31 5.27 6.79
Lower 95%
CL
140.23 -153.39 2.04 337.43 35.92 46.66 -803.69 -13.77 -16.19 -17.86
Upper 95%
CL
184.83 662.88 75.59 577.48 257.30 109.80 561.23 14.36 7.31 12.39
F1, 10 263.76 1.93 5.53 72.11 8.71 30.49 0.16 <0.01 0.71 0.16
P <0.001 0.195 0.041 <0.001 0.015 <0.001 0.701 0.963 0.419 0.696
Illness Mean -119.00 1085.85 804.52 -234.54 525.02 707.60 -267.01 -4.19 17.80 -18.88
SE 45.82 596.08 71.29 234.15 200.61 62.55 1076.20 25.28 19.08 26.67
Lower 95%
CL
-221.09 -242.30 645.67 -756.25 78.04 568.24 -2664.94 -60.52 -24.70 -78.31
Upper 95%
CL
-16.92 2414.01 963.37 287.18 972.00 846.97 2130.92 52.15 60.30 40.56
F1, 10 6.75 3.32 127.35 1.00 6.85 127.99 0.06 0.03 0.87 0.50
P 0.027 0.099 <0.001 0.340 0.026 <0.001 0.809 0.872 0.373 0.495
Skin wounds Mean 2.15 92.64 18.77 18.16 -74.41 56.61 1644.17 -0.43 -1.80 -0.90
SE 21.08 384.19 32.20 105.85 125.56 28.33 890.20 15.76 13.44 17.03
Lower 95%
CL
-44.82 -763.39 -52.99 -217.68 -354.18 -6.52 -339.32 -35.55 -31.75 -38.84
Upper 95%
CL
49.11 948.66 90.52 254.00 205.36 119.73 3627.65 34.70 28.14 37.04
F1, 10 0.01 0.06 0.34 0.03 0.35 3.99 3.41 <0.01 0.02 <0.01
P 0.921 0.814 0.573 0.867 0.567 0.074 0.095 0.979 0.896 0.959
Small Mean 148.00 -496.08 207.95 -5.35 312.85 237.95 227.03 2.77 -6.26 -6.93
SE 10.54 398.05 18.86 59.43 39.10 14.56 421.99 5.59 3.77 5.78
Lower 95%
CL
124.51 -1382.98 165.92 -137.75 225.72 205.51 -713.22 -9.69 -14.67 -19.81
Upper 95%
CL
171.49 390.82 249.98 127.06 398.98 270.40 1167.28.95 15.23 2.15 5.96
F1, 10 197.09 1.55 121.55 0.01 64.01 267.09 0.29 0.25 2.75 1.43
P <0.001 0.241 <0.001 0.930 <0.001 <0.001 0.602 0.631 0.128 0.259
Welfare problems index
Mean 118.19 79.18 98.53 243.75 266.67 135.85 184.61 0.78 -3.39 -6.051
SE 5.74 63.11 11.01 21.17 28.26 8.74 138.77 3.37 2.40 3.56
Lower 95%
CL
105.78 -57.15 74.75 198.01 205.63 116.96 -115.17 -6.51 -8.57 -13.75
Upper 95%
CL
130.60 215.51 122.31 289.49 327.72 154.73 484.40 8.06 1.79 1.65
F1, 13 423.44 1.57 80.14 132.53 89.07 241.48 1.77 0.05 2.00 2.88
P <0.001 0.232 <0.001 <0.001 <0.001 <0.001 0.206 0.822 0.181 0.113
Litter score Mean -0.16 1.47 0.36 1.62 0.59 0.50 -1.87 0.03 <0.01 0.02
SE 0.04 0.82 0.06 0.25 0.17 0.06 1.42 0.03 0.02 0.03
Lower 95%
CL
-0.24 -0.31 0.22 1.08 0.22 0.38 -4.93 -0.03 -0.05 -0.05
Upper 95%
CL
-0.08 3.25 0.50 2.16 0.96 0.62 1.19 0.08 0.05 0.08
F1, 13 18.25 3.19 31.70 42.61 12.05 76.19 1.75 0.93 <0.01 0.28
P <0.001 0.097 <0.001 <0.001 0.004 <0.001 0.209 0.353 0.988 0.605
1SeeMethodsandTable 3footnotes for explanation of variables.
2Walking difficulties, illness, skin wounds and small were predictors in one model, and the welfare problems index and litter score were predictors in separate models.
https://doi.org/10.1371/journal.pone.0214070.t006
